As extreme weather events become more frequent and severe because of climate change, investors need to reassess vulnerable assets, especially in renewable energy, where changing weather patterns directly affect performance. Improving modeling creates more accurate predictions, allowing investors to plan for both short-term and long-term effects.

To better understand how climate variability affects the returns of renewable energy assets, Milliman recently analyzed a major European solar farm using two models: a traditional internal rate of return (IRR) cash-flow model to assess long-term profitability, and a generalized additive model (GAM) to capture the complex, nonlinear effects of climate variables on electricity prices. By using both models together, the study examined how climate factors like temperature, solar radiation, and cloud cover influence revenue generation and the overall return on investment. It is important to note that this was a case study performed in Europe, and therefore results should be viewed within that regional context.

In this executive summary, we provide highlights of the study and commentary on the utility of the overall approach. To read the full study, which includes details of statistical methodologies, data preparation, and analytical results, download it here.

A dual approach to modeling

The IRR model is useful for assessing long-term profitability by providing a clear measure of expected returns based on projected cash flows. It helps investors compare different projects by calculating the rate at which an investment breaks even, making it ideal for more stable conditions over longer periods.

However, IRR faces challenges when dealing with nonlinear factors such as fluctuating electricity prices or changes in climate, as it assumes a steady, linear environment. In renewable energy sectors, where conditions can vary widely, this can lead to less accurate predictions.

The GAM is better at handling complex, nonlinear variables. It helps predict outcomes by looking at multiple factors, allowing each factor to have its own flexible curve, instead of assuming a straight-line relationship. This makes GAM useful for modeling the effects of weather patterns or energy prices, for example.

Compared to traditional regression models like multiple linear regression (MLR), GAM showed improved accuracy in this study, especially in handling seasonal and nonlinear patterns, with an R-squared of 72% for both the training and testing sets. In particular, the inclusion of hourly data helped capture daily energy price cycles, which improved the model's accuracy.

In this study, we used both types of modeling. GAM predicted electricity price changes based on climate data, while IRR assessed the long-term returns. By using GAM’s more realistic price projections in the IRR model, we gained a fuller understanding of how climate variability affects the solar farm’s profitability over time.

Key findings: Impact of climate variables on solar farm performance

The study highlighted several climate factors as having significant effects on electricity prices, which directly influenced the solar farm’s financial performance.

• Temperature: Higher temperatures had a modest but noticeable negative effect on electricity prices, likely due to lower demand for heating.
• Solar radiation: Increased solar radiation led to lower electricity prices because of the increased energy supply flowing into the grid.
• Cloud cover and precipitation: More cloud cover reduced solar energy production, which could limit supply and push prices up.

Despite these climate-related fluctuations, temperature increases had little impact on the long-term output of the solar farm in this particular region, where the rate of use of air conditioning is low.

Overall, while climate variability did affect short-term revenue through its impact on electricity prices, the long-term returns of the solar farm relied more on factors like the efficiency of the solar panels and the length of the investment period. The study examined different efficiency levels ( 11.35% and from 16% to 22%) and found that higher efficiency greatly improved returns over the 35-year period. Contrary to initial expectations, rising temperatures did not lead to higher prices or increased solar farm output cash-flows, showing that local energy demand patterns play a key role in long-term viability.

Broader applicability of the approach

This dual-model approach, combining the GAM and IRR, offers clear benefits for evaluating renewable energy assets. By using detailed climate data in the projections, this approach helps investors understand how changes in climate conditions—like solar radiation and temperature—affect electricity prices and revenue over time.

The study also challenged the initial expectation that rising temperatures would increase energy prices. In this region, temperature increases had little effect on prices, as demand for cooling is low. Instead, the efficiency of the solar panels and the fluctuations in electricity prices played a larger role in shaping long-term returns.

Several factors should be considered when applying this method:

• Regional differences: Energy use patterns and climate conditions differ from one area to another. It’s important not to assume that results will be the same everywhere for a given asset.
• Market dynamics: Energy markets vary in structure, and factors such as regulations, subsidies, and grid integration can impact both short-term revenue and long-term outcomes.
• Technology assumptions: Solar panel efficiency is critical to financial performance. As technology changes, it’s important to update assumptions about efficiency, depreciation, and operating costs to keep projections accurate.
• External risk factors: While GAM models climate-driven price changes well, external factors like geopolitical events or supply chain issues can create additional price volatility that historical data might not capture.

Overall, our study demonstrates that combining advanced models like GAM and IRR provides a more comprehensive understanding of the investment potential of renewable energy assets. This research highlights the importance of advanced predictive models for managing risks and identifying opportunities in renewable energy projects, supporting the broader conversation around sustainable investment strategies. These findings are consistent with other academic research in the field, further validating the model's application.

At Milliman, we continue to explore advanced modeling techniques to help clients navigate the complexities of renewable energy investments. By using data-driven models like GAM and IRR, we provide insights that support better decision making in today’s rapidly changing energy landscape.